Disturbance observer based active vibration suppression for force controllers [abstract]

Abstract

The lack of realistic haptic feedback poses a significant barrier to the full realization of haptic technology in robotics and virtual reality applications. Traditional model-based approaches, such as the spring-damper model, have relied on motion parameters and force sensors for capturing haptic sensations, despite their inherent drawbacks, including signal noise, narrow bandwidth, complexity, non-collocation, and instability. In contrast, sensorless force control mechanisms offer more accurate force measurements, which are crucial for proper object identification and accurate haptic object modeling. Recent research has focused on employing AI technologies for disturbance observer-based vibration suppression in motion controllers. AI approaches are data-driven, and a comprehensive understanding of the data is essential for optimal performance. Consequently, detailed statistical analysis is required to identify patterns, relationships, and significant features within the data. Using Disturbance Observer (DOB) and Reaction Force Observer (RFOB) based sensorless techniques, data was statistically analyzed to select important features for accurate object reconstruction and vibration suppression in motion controllers. A feature matrix was derived from mathematically defined features, and the performance of various AI algorithms was assessed by varying the feature input to the AI model. This comparison allowed for an understanding of the effect of the feature matrix on the model's performance. The bestfitting AI algorithm for the dataset, considering the selected feature matrix, was identified and trained for object reproduction and vibration suppression. This study introduces an AI-based approach for vivid force generation using a virtual model that replicates the actual environment with information derived from DOB and RFOB based sensorless force control mechanisms with active vibration suppression. The validity of the AI approach was evaluated by comparing the performance of the AI model with the conventional object model and the actual object, demonstrating the potential for improved vibration suppression in motion controllers.